Review of Computer Engineering Research
December 2016, Volume 3, 4, pp 6981
Shoewu, O, Salau, N.O, Ogunlewe, A.O, Oborkhale, L.I
Article History:
Received:
16 November, 2016
Revised:
20 December, 2016
Accepted:
02 January, 2017
Published:
20 January, 2017
This paper presents path loss measurement and modeling for Lagos state denseurban (DU), urban (UR), suburban (SU) and nonurban (NU) G.S.M environments. It was carried out with data collection through drive testing using TEMS software in the chosen environments LagosIsland(DU), Surulere (UR), LekkiOniru (SU), AgbedeIkorodu (NU), over a distance of 0.510Km from Base station (BS) to Mobile station (MS) with measurement taken at 0.5Km intervals for a period of 52 weeks. Relative parameters like Rxlev, RSSI, Path loss were measured in all areas of investigation under 2G and 3G frequencies of operation and twelve (12) different sites location were covered and analysed. COST 231Hata model was used as reference model for path loss calculation of field data, this was further adjusted to develop optimized models (tagged OMODEEN models) for path loss prediction in all environments of study, which shows results within 6dB acceptable range, hence recommended for modeling in these environs and other similar G.S.M environments.
Keywords: GSM, Modeling, Environment, KPI and path loss.
Received: 16 November 2016/ Revised: 20 December 2016/ Accepted: 2 January 2017/ Published: 20 January 2017
This study is one of very few studies which have investigated path loss in classified environment. Effort was geared towards four chosen environment in Lagos metropolitan terrain. A new model known as OMODEEN was developed and can be used for specific environs. MATLAB was used to develop graphical representation of the model.
Since the essential goal of any global system for mobile communication (GSM) service provider is to provide excellent services to her subscribers, which might be impeded by many effects like reflection, refraction, diffraction, scattering and absorption, which introduce path loss to radio communication between the base transceiver station (BTS) of the provider and the mobile unit (MU) with the subscriber, it becomes imperative to constantly investigate and model this path loss which is also influenced by terrain contours, environment (urban or rural, vegetation and foliage), propagation medium (dry or moist air), the distance between the transmitter and the receiver, and the height and location of antennas. Therefore this work presents path loss measurement and modeling for Lagos state denseurban, urban, suburban and nonurban, G.S.M environments.
In this research work, COST 231 model was used to predict path loss over the range of distance covered and further adjusted by finding the RMSE values between measured path loss and COST 231 predicted path loss to obtain optimized path loss prediction model tagged (OMODEEN).
COST 231Hata model was chosen as a reference model because of its peculiarity which makes it useful for predicting signal strength in all environments [1, 2] its frequency range that extends to 2000 MHz [3] and its incorporated signal strength prediction of up to 20km from transmitter to receiver with transmitter antenna height ranging from 30m to 200m and receiver antenna height ranging from 1m to 10m [1, 4].
Past projects [5, 6] have also suggested the Cost 231Hata model to show the best performance in Lagos environments, hence its adoption as a reference model.
While the COST 231 showed satisfactory RMSE values of 3.23dB, 1.7dB, 3.88dB,7.11dB for 2G900MHz, 3.33dB, 3.38dB, 6.15dB, 6.49dB for 2G1800MHz, and 5.08dB, 2.99dB, 5.70dB, 9.24dB for 3G2100MHz, in nonurban, suburban, urban and denseurban environments respectively, this model when modified (OMODEEN) was found to predict path loss better with RMSE values of 2.70dB, 1.60dB, 3.12dB, 5.62dB for 2G900MHz, 2.66dB, 2.70dB, 4.83dB, 5.08 dB for 2G1800MHz, and 4.04dB, 2.43dB, 4.48dB, 6.25dB for 3G2100MHz, in nonurban, suburban, urban and denseurban environments respectively, which are acceptable for prediction purposes.
Efforts have been exerted on measurement and analysis of a particular terrain e.g. SubUrban [5] or path loss modeling of three different environs using just one frequency of modulation (1800MHz) [6] we therefore tried to take these studies further by considering four different G.S.M environments, modeled for these environments putting the 2G and 3G frequencies, i.e. 900MHz, 1800MHz, and 2100MHz into research consideration.
Fig1. Map of Lagos State from Google Map
Table1. GPS Values From Measured Data
SN 
Environments of study with G.P.S Values 

Environment 
Location 
G.P.S 

1 
Non Urban 
AgbedeIkorodu 
N6^{0} 39.9250' E3^{0} 29.0363' 
2 
Sub Urban 
LekkiOniru 
N6^{0} 26.6661' E3^{0} 28.7463' 
3 
Urban 
Surulere 
N6^{0} 33.3844' E3^{0} 20.9407' 
4 
Dense Urban 
LagosIsland 
N6^{0} 27.4832' E3^{0} 23.5453' 
Fig2. TEMS Investigation Interface for NonUrban(AgbedeIkorodu) Environ from logfiles
Fig3. TEMS Investigation Interface for SubUrban(LekkiOniru) Environ from Mapinfo
Fig4. TEMS Investigation Interface for Urban(Surulere) Environ from Mapinfo
Fig5. TEMS Investigation Interface for DenseUrban(LagosIsland) Environ from Mapinfo
A. Measured Parameters In 2G Environments
Quality of GSM/EDGE 2G 900/1800 MHz coverage is described basically by two indicators (KPIs); according to ECC Report 118 (2008) [7] these are:
Receive Level
RxLev –this is the received signal strength on serving cell, measured respectively on all slots RxLevFull and on a subset of slots RxLevSub. RxLevel is received power level at MS (maximum RxLevel measured by MS is (±) – 40 dBm [8].
Receive Quality
B. Measured Parameters in 3G Environments
Quality of UMTS 3G 2100 MHz coverage is described basically by three indicators, according to the ECC Report 103 (2007) [7] these are
RSCP is the received power on one code measured on the pilot bits of the PCPICH (Primary Common Pilot Channel).
RSSI is the wideb and received power within the relevant channel bandwidth; it is the measure of received signal strength in 3G domain.
Ec/N0 is the ratio of received pilot energy, Ec, to the total received energy or the total power spectral density, I0 .The received energy per chip, Ec, divided by the power density in the band. The Ec/N0 is identical to RSCP/RSSI [7]. Measured in decibel; dB [5].
C. Methodology
Algorithm of Research Study
STEP 1: Drive Tests using TEMS 13 were carried out in four G.S.M environments in Lagos, data were analysed with Mapinfo 11 and extracted into Excel format.
STEP 2: RxLev (2G) and RSSI (3G) from 12 Base stations in total were recorded, Measurements range between BTS and MS is 0.5 to 10Km apart.
STEP 3: Measurements were taken at intervals of 0.5 Km twice in a day in all environments and mean values calculated over a period of 52 weeks.
STEP 4: AgbedeIkorodu, LekkiOniru, Surulere and Lagos Isl and were chosen as NonUrban, Sub Urban, Urban and DenseUrban environment respectively.
STEP 5: Path loss measured were compiled using equation (1)
STEP 6: Calculated (empirical) path loss were compiled using COST 231 equation (3)
STEP 7: RMSE of Calculated (COST 231) path loss and measured path loss were found, using statistical formula equation (4)
STEP 8: The RMSE of calculated (COST 231) was used to modify the original COST 231 equation (3) to obtain new model referred to as Optimized model.
STEP 9: The optimized model was used to calculate new PREDICTED path loss values.
STEP 10: RMSE of new predicted path loss values (Optimized model) and measured path loss were found, and compared with RMSE at STEP 7.
STEP 11: RMSE of STEP 10 was found to be of lesser values and also lower than 6dB standard, showing a better prediction and hence recommended for modeling
D. Experimental Setup of Drive Test
Fig6. Drive Test TEMS Phones of the Experimental setup
Fig7. Flow chart of Drive Test for the Study
E. Data Analysis of Measured Pathloss
Table2. G.S.M Environments and R.F Parameters
Environment 
BTS Power 
BTS Antenna Height 
Non urban(Rural) 
43dBm 
45m 
Urban 
38 dBm 
35m 
Suburban 
43 dBm 
40m 
DenseUrban 
36 dBm 
30m 
Where Connection loss = 4.3dBi, Feeder loss=0.3dBi, Duplexer loss=2.1dBi, Antenna Gain=2.1dBi, BTS antenna Gain=14dBi
The measured path loss PL_{m} (dB) for each measurement location at a distance d(km) can be found by equations given by Rappaport [9] and Seybold [10] as:
PL_{m}(dB)=EIRP_{t}(dBm)–P_{r}(dBm)………………,...,,,,,,,.(1)
Where EIRP_{t} = effective isotropic radiated power in dBm and P_{r} = mean power received in dBm.
The effective isotropic radiated power EIRP_{t}(dBm) is given as:
EIRP_{t}=P_{BTS}+G_{BTS}+G_{MS}–L_{FC}–L_{AB}–L_{CF}……………..... (2)
Where P_{BTS} = BTS power (dBm),
G_{BTS} = BTS antenna gain (dBi), G_{MS}=MS antenna gain (dBi),
L_{FC} = feeder cable and connector loss (dB),
L_{AB} = antenna body loss (dB) and L_{CF} = combiner and filter loss (dB).
Substituting the values in Table II into equation (2), we calculated EIRP_{t} , the EIRP_{t} values calculated above were further inserted into equation (1) and the tables for Path loss measured (PL_{m}) were compiled.
4.1. Data Analysis of Calculated (COST 231) Path loss
Calculations of Empirical Path loss were achieved using Cost 231 path loss model equation
P_{L}(dB)=46.3+33.9log_{10}f_{c}−13.82log_{10}(h_{t})−a(h_{r})+[44.9−6.55log10 h_{t}]log_{d0} d + c .........................................................(3)
Where:
C= 0 dB, for suburban areas or open environments and 3dB for Urban environment [5, 6]
(h_{r})= mobile station antenna height correction factor is defined as
a(h_{r})=(1.11log_{10}f_{c}−0.7) h_{r}−(1.5log10f_{c}−0.8) , for suburban or rural areas [5, 6]
a(h_{r})=3.20[log_{10}(11.75hr)]  4.97 for f > 400MHz for Urban environment [6]
4.2. Data Analysis of RMSE
RMSE (Root mean square error) statistic gives a quantitative measure of how close the predicted path loss values (COST 231) are to the measured path loss values. RMSE value closer to zero indicates a better fit. It is given as stated below
………… (4)
Where PLm (d) = measured path loss (dB), PLr (d) = calculated path loss (dB) and k = 20 (number of measured data points).
Equation (4) above was applied to the numerical values of the measured path loss and the predicted path loss on the basis of each propagation model to obtain the RMSEs for different environments under study as shown in Table III below.
Table3. Root Mean Square Error of Calculated Path loss and Measured Path loss
2G – 900 MHz 
2G–1800 MHz 
3G–2100 MHz 


NU 
3.23 
3.33 
5.08 
SU 
1.74 
3.38 
2.99 
UR 
3.88 
6.15 
5.70 
DU 
7.11 
6.49 
9.24 
The RMSE of calculated (COST 231) was used to modify the original COST 231 equation (3) to obtain new model referred to as Optimized model, by simply subtracting the RMSE values from the constant (46.3) value in the formula below
P_{L}(dB)=46.3+33.9log_{10}f_{c}−13.82log_{10}(h_{t})−a(h_{r})+[44.9−6.55log10h_{t}]log_{10} d+C
Table4. Residual values from COST 231 formula
2G – 900 MHz 
2G–1800 MHz 
3G–2100 MHz 


NU 
43.07 
42.97 
41.22 
SU 
44.56 
42.92 
43.31 
UR 
42.42 
40.15 
40.6 
DU 
39.19 
39.81 
37.06 
The optimized model taking care of the environments where tests were carried out will now have the following COST 231 modified equations as PREDICTION MODELS: named OMODEEN Path loss Prediction Model.
FOR 2G900 MHz
P_{L(NU)}=43.07+33.9log_{10}f_{c}−13.82log_{10}(ht)−a(hr)+[44.9−6.55log10ht]log_{10}d + C.........................(5a)
P_{L(SU)}=44.56+33.9log_{10}fc−13.82log_{10}(ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C.......................(5b)
P_{L(UR)}=42.42+33.9log_{10}fc−13.82log_{10}(ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C........................(5c)
P_{L(DU)}=39.19+33.9log_{10}fc−13.82log_{10}(ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C........................(5d)
FOR 2G1800 MHz
P_{L(NU)}=42.97+33.9log_{10}fc−13.82log_{10} (ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C..................(6a)
P_{L(SU)}=42.92+33.9log_{10}fc−13.82log_{10} (ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C..................(6b)
P_{L(UR)}=40.15+33.9log_{10}fc−13.82log_{10} (ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C..................(6c)
P_{L(DU)}=39.81+33.9log_{10}fc−13.82log_{10} (ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C..................(6d)
FOR 3G2100 MHz
P_{L(NU)}=41.22+33.9log_{10}fc−13.82log_{10}(ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C....................................................(7a)
P_{L(SU)}=43.31+33.9log_{10}fc−13.82log_{10}(ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C...................................................(7b)
P_{L(UR)}=40.60+33.9log_{10}fc−13.82log_{10}(ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C...................................................(7c)
P_{L(DU)}=37.06+33.9log_{10}fc−13.82log_{10}(ht)−a(hr)+[44.9−6.55log10 ht]log_{10} d + C...................................................(7d)
Substituting the f_{c} as appropriate  900, 1800 and 2100 and a(h_{r}) as given above and h_{t} from Table II above, where h_{r} (Height of MS)=3m, then we have simplified forms of OMODEEN Path loss Prediction Models as shown in equations (819) below, where d is distance between BTS and MS.(0.510Km).
FOR 2G900 MHz
P_{L(NU)}=116.53+34.07log_{10}d............................................(8)
P_{L(SU)}=118.73+34.41log_{10}d............................................(9)
P_{L(UR)}=118.54+37.79log_{10} d..........................................(10)
P_{L(DU)}=116.24+38.22log_{10} d..........................................(11)
FOR 2G1800 MHz
P_{L(NU)}=126.11+34.07log_{10} d........................................(12)
P_{L(SU)}=126.77+34.41log_{10} d.........................................(13)
P_{L(UR)}=126.47+37.79log_{10} d........................................(14)
P_{L(DU)}=127.06+38.22log_{10} d........................................(15)
FOR 3G2100 MHz
P_{L(NU)}=126.52+34.07log_{10} d........................................(16)
P_{L(SU)}=129.31+34.41log_{10} d........................................(17)
P_{L(UR)}=129.19+37.79log_{10} d........................................(18)
P_{L(DU)}=126.58+38.22log_{10} d........................................(19)
The optimized models, equations (819) were used to calculate new PREDICTED path loss values, and RMSE equation (4) was used to analyse its values with measured path loss to obtain the following table:
Table5. RMSE of Optimized Path loss Model from Measured Data
2G–900 MHz 
2G–1800 MHz 
3G–2100 MHz 

NU 
2.70 
2.66 
4.04 
SU 
1.60 
2.70 
2.43 
UR 
3.12 
4.83 
4.48 
DU 
5.62 
5.08 
6.25 
F. Result Analysis
Fig8. Bar chart of RMSE of Cost231 and Optimized Models from Excel
RMSE of optimized model found to be of lesser values and also within 6dB standard [11] virtually in all showing a better prediction and hence recommended for modelling.
We further used MATLAB 2015 edition to plot the graphs of the measured, calculated and optimized path loss values in all environments to test the correctness of our prediction models, hence we have:
Fig9. Matlab Plots of Measured, Cost231 and Optimized Path loss in 2G900MHz NonUrban environment
Fig10. Matlab Plots of Measured, Cost231 and Optimized Path loss in 2G900MHz SubUrban environment
Fig11. Matlab Plots of Measured, Cost231 and Optimized Path loss in 2G900MHz Urban environment
Fig12. Matlab Plots of Measured, Cost231 and Optimized Path loss in 2G900MHz DenseUrban environment
Fig13. Matlab Plots of Measured, Cost231 and Optimized Path loss in 2G1800MHz NonUrban environment
Fig14. Matlab Plots of Measured, Cost231 and Optimized Path loss in 2G1800MHz NonUrban environment
Fig15. Matlab Plots of Measured, Cost231 and Optimized Path loss in 2G1800MHz Urban environment
Fig16. Matlab Plots of Measured, Cost231 and Optimized Path loss in 2G1800MHz DenseUrban environment
Fig17. Matlab Plots of Measured, Cost231 and Optimized Path loss in 3G2100MHz NonUrban environment
Fig18. Matlab Plots of Measured, Cost231 and Optimized Path loss in 3G2100MHz SubUrban environment
Fig19. Matlab Plots of Measured, Cost231 and Optimized Path loss in 3G2100MHz Urban environment
Fig20. Matlab Plots of Measured, Cost231 and Optimized Path loss in 3G2100MHz DenseUrban environment
The results of this study revealed that the Cost 231Hata model showed a satisfactory performance in the chosen environments based on its RMSE values as shown in Table III, Path loss plots among Measured, Predicted (Cost231) and Optimized models as we have in figures (920) revealed the closeness of our optimized model results to the measured path loss, which shows accuracy of our results.
Likewise from Table IV and figure 8, it was observed that the RMSE values obtained from the optimized model is lower than the one from the predicting model (Cost 231), and at the same time meet the 6dB [11] standard, hence our prediction models, tagged OMODEEN can be used in these environments of study and in any other environments with similar characteristics.
Funding: This study received no specific financial support. 
Competing Interests: The authors declare that they have no competing interests. 
Contributors/Acknowledgement: Wish to appreciate Dr. Femi Ipinnimo, Dr Sikiru A Adegoke, staffs of the department, staffs of OMNICOM, and my excellent course mates for their wonderful contributions to make this research a reality 
[1] J. Isabona, C. C. Konyeha, B. C. Chinule, and G. P. Isaiah, "Radio field strength propagation data and pathloss calculation methods in UMTS network," Advances in Physics Theories and Applications, vol. 21, pp. 54  68, 2013.
[2] A. Goldsmith, Wireless communication. New York: Cambridge University Press, 2005.
[3] O. Shoewu, S. O. Olatinwo, and D. D. Olatinwo, "Comparative analysis of path loss in metropolitan domain in Nigeria using OkumuraHata model," European Journal of Scientific Research, vol. 4, pp. 5 15, 2016.
[4] R. Mardeni and T. S. Priya, "Optimized COST 231 hata models for WiMAX pathloss prediction in suburban and open urban environments," Modern Applied Science, vol. 4, pp.75, 2010. View at Google Scholar  View at Publisher
[5] L. Akinyemi, N. Makanjuola, O. Shoewu, and F. Edeko, "Evaluation and analysis of 3G network in Lagos metropolis Nigeria," International Transaction of Electrical and Computer Engineers System, vol. 2, pp. 8187, 2014. View at Google Scholar
[6] S. Ajose and A. L. Imoize, "Propagation measurements and modelling at 1800 MHz in Lagos Nigeria," International Journal of Wireless and Mobile Computing, vol. 6, pp. 165174, 2013. View at Google Scholar  View at Publisher
[7] I. U. Nochiri, C. C. Osuagwu, and K. C. Okafor, "Empirical analysis on the GSM network KPIs using realtime methodology for a novel network integration," Progress in Science and Engineering Research Journal, vol. 13, pp. 092107, 2014.
[8] O. Shoewu and F. O. Edeko, "Propagation loss determination in cluster based GSM base stations in Lagos environs," International Transaction of Electrical and Computer Engineers System, vol. 2, pp. 28  33, 2014. View at Google Scholar
[9] S. T. Rappaport, Wireless communications principles and practice, 2nd ed. U.S.A: University of Texas, 2002.
[10] S. J. Seybold, Wireless communication principles and practice, 2nd ed. U.S.A: Prentice Hall, 2005.
[11] T. Syahfrizal, "Using of the OkumuraHata propagation model for pathloss determination in Tarakan," in Proceedings of the 3rd International Conference on Radar, Antenna, Microwave, Electronics and Telecommunications (ICRAMET), 2014, pp. 7779.
Views and opinions expressed in this article are the views and opinions of the author(s), Review of Computer Engineering Research shall not be responsible or answerable for any loss, damage or liability etc. caused in relation to/arising out of the use of the content. 